How much VRAM does Solar-Open2-250B need?

Q4_K_M is exactly 151,819,124,128 bytes (141.39 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Solar-Open2-250B's KV cache?

6.00 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

Is Solar-Open2-250B a mixture-of-experts model?

Yes — 320 experts with 8 routed per token. Every expert must be resident in memory, but only the routed ones are read per token, so its memory requirement and its speed behave very differently.

upstage · text · mixture of experts

Solar-Open2-250B

upstage/Solar-Open2-250B

Solar-Open2-250B at Q4_K_M is exactly 151,819,124,128 bytes (141.39 GiB / 151.82 GB) — an effective 4.853 bits per weight, not the nominal 4. Its KV cache at 32K is 6.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

250B

total, not active

Architecture

solar-open2

48 layers

Context

1,048,576

native (config.json)

License

other

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_M	52.67 GiB	56,552,373,920	1.808	—	vcruz305
IQ1_M2 shards	56.30 GiB	60,450,717,632	1.932	—	prometheusAIR
IQ2_M	76.24 GiB	81,861,402,272	2.617	—	vcruz305
Q2_K	85.18 GiB	91,458,059,680	2.923	—	vcruz305
Q2_K3 shards	88.87 GiB	95,427,444,768	3.050	—	prometheusAIR
IQ3_XXS	89.86 GiB	96,483,974,816	3.084	—	vcruz305
Q3_K_M	111.63 GiB	119,864,450,720	3.831	—	vcruz305
IQ4_XS	124.46 GiB	133,638,880,928	4.271	—	vcruz305
IQ4_XS4 shards	126.88 GiB	136,236,466,304	4.355	—	prometheusAIR
Q4_K_M	141.39 GiB	151,819,124,128	4.853	—	vcruz305
Q5_K_M	165.65 GiB	177,863,785,888	5.685	—	vcruz305
Q6_K5 shards	191.44 GiB	205,554,819,072	6.570	—	prometheusAIR
Q8_0	247.86 GiB	266,136,845,728	8.507	—	vcruz305

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.75 GiB	0.75 GiB	—	48 / 0 / 0
8,192	1.50 GiB	1.50 GiB	—	48 / 0 / 0
16,384	3.00 GiB	3.00 GiB	—	48 / 0 / 0
32,768	6.00 GiB	6.00 GiB	—	48 / 0 / 0
65,536	12.00 GiB	12.00 GiB	—	48 / 0 / 0
131,072	24.00 GiB	24.00 GiB	—	48 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-Coder-30B-A3B-Instruct vs Qwen3.6-27B vs Qwen3.8-27B vs DeepSeek-V4-Flash vs Qwen3-30B-A3B-Thinking-2507 vs Laguna-XS-2.1 All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 131.12 GiB. The real file is 141.39 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

128

Hidden size

4096

Vocab

196,608

Sliding window

none

SWA period

—

MLA

Experts

320

Experts per token

use_sliding_window

—

Questions people ask

How much VRAM does Solar-Open2-250B need?: Q4_K_M is exactly 151,819,124,128 bytes (141.39 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Solar-Open2-250B's KV cache?: 6.00 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Is Solar-Open2-250B a mixture-of-experts model?: Yes — 320 experts, 8 routed per token. Every expert must be resident, but only the routed ones are read per token, which is why its memory requirement and its speed behave very differently.
Which quantization of Solar-Open2-250B should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.